How to diagnose crash

[jeffreywhunter]

I've got a 6.3.5 server that runs fine for a couple days, then stops communicating.  No login from web or console.  Only recourse is to do a hard reset, which is not a happy situation.  I'm looking for advice on how to diagnose this.  

 

I take a look at the log all the time and dont see anything that looks ominous (except I am having a problem with user inotify watches being exceeded (even though set to 720,000).  See that discussion here: https://forums.lime-technology.com/topic/58894-still-not-enough-inotify-watches/

 

I've disabled dockers and no change.  Is there a way to dump the log when the system becomes unresponsive?  or is there a way to have the log written to a file so we can get an inkling of what is going on?  Maybe we'd see something in the last bit of the log that would indicate something...

 

Thanks in advance...

[Squid]

In no particular order, 

 

• Check for bios updates
• Run memtest
• Disable any VMs that utilize hardware passthrough
• Power Supply Issues
• Mains voltage dipping and UPS not kicking in or no UPS present.  (ie: do your lights dim in the house when your fridge kicks in?) - Ideally servers should be on a separate isolated electric line, and if not feasible, then definitely not on the same line as anything that has a motor in it
• CPU overheating
• Hook up a monitor to the server and see what's on its display at time of non-responsivenes

Log saved up to crash:  Either Fix Common Problems in troubleshooting mode (will also log a ton of other info that may or may not be of interest) or from a command line (ideally via screen or the actual local monitor / keyboard)

tail -f /var/log/syslog /boot/syslog.txt

 

[JonathanM]

13 hours ago, Squid said:

definitely not on the same line as anything that has a motor in it

or laser printer.